fix: use fast-array-utils’ threadsafe njit

[flying-sheep]

• Closes #
• Tests included or not required because:

• Release notes not necessary because:

@ilan-gold ported scanpy’s njit wrapper to fast-array-utils where I finally fixed it so it works as intended

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77.84%. Comparing base (fe558a9) to head (08816f1).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4015      +/-   ##
==========================================
- Coverage   78.33%   77.84%   -0.50%     
==========================================
  Files         117      117              
  Lines       12769    12740      -29     
==========================================
- Hits        10003     9917      -86     
- Misses       2766     2823      +57     

Flag	Coverage Δ
hatch-test.low-vers	77.79% <100.00%> (+0.08%)	⬆️
hatch-test.pre	20.55% <100.00%> (-48.71%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/scanpy/_compat.py	83.67% <100.00%> (+12.74%)	⬆️
...c/scanpy/experimental/pp/_highly_variable_genes.py	94.23% <100.00%> (+0.05%)	⬆️
src/scanpy/metrics/_gearys_c.py	100.00% <100.00%> (ø)
src/scanpy/metrics/_morans_i.py	100.00% <100.00%> (ø)
src/scanpy/preprocessing/_normalization.py	94.66% <100.00%> (+0.07%)	⬆️
src/scanpy/preprocessing/_qc.py	94.48% <100.00%> (-1.36%)	⬇️
src/scanpy/preprocessing/_scale.py	89.58% <100.00%> (+0.10%)	⬆️
src/scanpy/preprocessing/_simple.py	90.99% <100.00%> (+0.02%)	⬆️
src/scanpy/tools/_rank_genes_groups.py	92.83% <100.00%> (+0.02%)	⬆️

... and 8 files with indirect coverage changes

[ilan-gold]

Looks we still got an old (broken) import in there judging by the tests!

[ilan-gold]

I added the benchmark label. As long as nothing regresses horribly, I think we're good

[scverse-benchmark]

Benchmark changes

Change	Before [2ddfcc7]	After [f07b885]	Ratio	Benchmark (Parameter)
-	482±40μs	388±2μs	0.81	preprocessing_counts.FastSuite.time_log1p('pbmc68k_reduced', 'counts-off-axis')

Comparison: https://github.com/scverse/scanpy/compare/2ddfcc74ecc3d47124ca25848d8ea9f97ddd7f83..f07b885f9bda5480567971ec71e97831e191bf50
Last changed: Tue, 31 Mar 2026 08:57:44 +0000

More details: https://github.com/scverse/scanpy/pull/4015/checks?check_run_id=69314496996